Double Jeopardy and Climate Impact in the Use of Large Language Models: Socio-economic Disparities and Reduced Utility for Non-English Speakers
This research addresses socio-economic disparities in AI access, revealing how tokenization in LLMs exacerbates inequalities for non-English speakers, particularly in developing nations, making it a critical issue for fair algorithm development.
The study found that large language models (LLMs) disproportionately disadvantage non-English speakers, with users in low-income countries facing 4 to 6 times higher costs and poorer performance in translation tasks, highlighting a 'double jeopardy' of increased expenses and reduced utility.
Artificial Intelligence (AI), particularly large language models (LLMs), holds the potential to bridge language and information gaps, which can benefit the economies of developing nations. However, our analysis of FLORES-200, FLORES+, Ethnologue, and World Development Indicators data reveals that these benefits largely favor English speakers. Speakers of languages in low-income and lower-middle-income countries face higher costs when using OpenAI's GPT models via APIs because of how the system processes the input -- tokenization. Around 1.5 billion people, speaking languages primarily from lower-middle-income countries, could incur costs that are 4 to 6 times higher than those faced by English speakers. Disparities in LLM performance are significant, and tokenization in models priced per token amplifies inequalities in access, cost, and utility. Moreover, using the quality of translation tasks as a proxy measure, we show that LLMs perform poorly in low-resource languages, presenting a ``double jeopardy" of higher costs and poor performance for these users. We also discuss the direct impact of fragmentation in tokenizing low-resource languages on climate. This underscores the need for fairer algorithm development to benefit all linguistic groups.